Data for "Application of Machine Learning Techniques to Map Wetland Types in the Numto Nature Park (Western Siberia)"

Python code

The numto.ipynb file is the Jupyter Notebook with Python code that was used for data preprocessing, training the models and mapping the modeling results.

Model checkpoints

This repository contains only joblib dumps of trained classical machine learing models. Model checkpoints for deep learning models are too large for GitHub and can be only provided by request.

File	Model
lr.joblib	Linear regression
rf.joblib	Sklearn Random Forest
xgbrf.joblib	XGBoost Random Forest
xgb.joblib	XGBoost Gradient Boosting

Maps

File	Map
map_nn.tif	Modeling with DeepLabV3
map_nn_clip.tif	Same, but clipped
map_nn_train.tif	Same, but only for park area
map_nn_train_clip.tif	Same, but only for park area and clipped
map_xgb.tif	Modeling with Gradient Boosting
map_xgb_train.tif	Same, but only for park area

Map legend:

Value	Class
0	No data
1	Lakes
2	Forests
3	Burnt areas
4	Palsa mires
5	Oligotrophic mires
6	Eutrophic and mesotrophic mires
7	Built-up areas